Data Needs for the State Children's Health Insurance Program (2002)

Collaboration among state SCHIP programs is an important way to expand knowledge about effective ways of, and problems in, administering SCHIP programs. The workshop itself was evidence of this. More generally, a number of issues common to most SCHIP programs would be informed by cross-fertilization among the states. The development of common approaches could benefit many of the state programs.

Estimating eligibility, enrollment, disenrollment, and more generally, understanding how children (and families) move among the different insurance categories—SCHIP, Medicaid, private insurance, or no insurance at all—are common goals of the SCHIP programs across the states. Each state has developed its own methods of measurement to deal with these issues, sometimes adopting methods used by one or more other states. Central development and dissemination of analytic methods to accomplish these goals would be useful to the individual SCHIP programs.

Another common issue is the tremendous movement among insurance statuses and the problem of how to measure these changes. As indicated earlier, 27 states conduct their own surveys to attempt to measure the extent of these transitions. Dubay described the Urban Institute’s National Survey of America’s Families, which provides a national picture and has a

sample size large enough to provide detailed data for 13 states. It is difficult for many states to adopt this methodology, primarily for budgetary reasons. Norton’s experience in New Hampshire is evidence of this; his budget allowed him to use only a stripped-down version of this survey.

The State Health Access Data Assistance Center at the University of Minnesota has been organized to provide assistance to states in dealing with these issues. The center is funded by a grant from the Robert Wood Johnson Foundation to provide technical assistance to states that are interested in collecting relevant data for state health policy. Lynn Blewett described what states want from national surveys. The list includes data that are representative of the individual state; a sample size that is large enough to provide valid and reliable estimates; a survey design that produces policy-relevant information; timely and routine release of data; and access to micro data for further state-specific analyses.

The effect of missing data in surveys can be a critical issue in deriving estimates from surveys. Panel member Paul Newacheck pointed out that questions on income, for example, are known to have large nonresponse rates in some of the major national surveys. He raised the question of how much imputation is going into the microsimulation models that produce estimates on insurance eligibility and what effect this might have on the estimates. If as much as a quarter of the data are imputed, this could have a substantial effect on the validity of the analysis. The discussion that followed provided no direct answer to the question, but some of the presenters stated that when they attempted to compare survey results, they found considerable similarity. This seemed to give them confidence in the use of the data from the surveys, despite high levels of missing data.

Several of the workshop participants pointed out that data definitions are not, in general, standard across states or across databases of related federal programs. The criteria for eligibility for SCHIP vary considerably among the states, not only due to the differing income limits for SCHIP and Medicaid among the states, but also due to how income is defined for determining eligibility. Perhaps the most detailed set of factors for determining income is that used in the National Survey of America’s Families. At the other end of the scale, several states use self-declaration of income to determine eligibility with considerable variation in the extent of the guidelines that are given the applicant on what to include as income. Superim

posed upon this is the fact that the employment status of the low-income population is so volatile that income may change several times during a short period, altering a given individual’s eligibility status several times. Ellwood proposed that demographic data as well as reasons for disenrollment be reported in standardized categories to facilitate relating outcomes to underlying factors across the states.

In order to model how children move among the health insurance statuses, it would be helpful to have a consistent family identifier in SCHIP and Medicaid data sets so that the health insurance status of children from the same family can be tracked. Also, matching information from different datasets is a valuable way to identify who is eligible for coverage but not using SCHIP. The Medical Statistical Information System of the Centers for Medicare and Medicaid Services, described above, is a mechanism that can shed light on this issue. It has the advantage of giving children unique identifiers (Social Security numbers for most states) so that tracking them over time as they change statuses and tracking children from the same family are possible. Data are not yet available from this system, but there are preliminary indications of major differences among states in turnover and in transfers between Medicaid and SCHIP as well as the extent of short-term gaps in enrollment. It should be pointed out, however, that computer matching of records is not without its problems. Robert Gellman cautioned that the Computer Matching and Privacy Protection Act of 1988 limits the extent to which computer matching of records is permitted if at least one of the datasets contains federal records. This act lists a set of requirements that must be met to allow matching to proceed. Gellman pointed out that compliance with this act and enforcement of its provisions have been mixed, but that those who plan to conduct a match involving federal records should be aware of the provisions and take them into account.

In December 2000, the Office of the Assistant Secretary for Planning and Evaluation of the Department of Health and Human services issued a contract for a congressionally mandated study to evaluate the impact of the SCHIP (Mathematica Policy Research, 2002). The study was to include

10 states that would have some degree of national representation. Although the study places substantial emphasis on such program aspects as outreach, enrollment efforts, relation to Medicaid and private insurance, etc., it also includes plans to address access to health care and utilization of health services. With respect to those issues, the researchers were asked to pursue the following questions:

• What experiences do SCHIP enrollees have in seeking and obtaining services, and how does this compare with their experiences prior to enrollment?

• What proportion of SCHIP enrollees has a usual source of medical care?

• How does the program and benefit design impact access and utilization of services?

• How satisfied are enrollees with SCHIP and the health services they receive through the program?

• How adequate are states’ or contracted provider networks in meeting the need of SCHIP enrollees?

Data that bear on questions of this type would be extremely helpful to each state for monitoring the effectiveness of its SCHIP.

In evaluating SCHIP, it is important to use qualitative as well as quantitative information. As the workshop discussions reflected, very little information has been available to indicate reasons for the failure of eligible children to enroll in SCHIP or the failure of those in the program to reenroll when the time for renewal arrives. Most of the national and state surveys are not designed to provide this information. Qualitative information can be quite useful for this purpose. The focus groups, mentioned above, that Bellamy will be conducting are aimed at obtaining answers from four different groups: families who are eligible for either Medicaid or SCHIP but not enrolled, families who are enrolled, families who are disenrolled, and families who have private insurance. The discussions in each of the groups will be focused on questions aimed at uncovering what the participants know about health insurance coverage for children. This is the kind of information that will be helpful to those administering SCHIP and Medicaid in making their programs more accessible.

On these cross-cutting issues, the panel drew the following conclusions:

• Federal-state cooperation is essential in developing a national strategy to disseminate “best data practices” across states.

• The use of family identifiers on a sample basis for modeling, within the constraints of privacy considerations, is a tactic to be considered to match insurance coverage data from different datasets and to track health insurance coverage for children in the same family.

• An appropriate federal or private national agency should undertake the following:

  • Develop a central repository for analytic expertise on methods for conducting sample audits.

  • Disseminate analytical models for handling missing data and survey nonresponses in statistical modeling.

  • Develop and share among the states protocols for obtaining qualitative information to assess reasons for lack of enrollment and for disenrollment.